Use NEON reg-reg moves, where profitable. This reduces "domain-cross" stalls, when we used to mix vfp and neon code (the former were used for reg-reg moves)

llvm-svn: 85764
diff --git a/llvm/lib/Target/ARM/ARMInstrInfo.cpp b/llvm/lib/Target/ARM/ARMInstrInfo.cpp
index c24c690..86bbe2a 100644
--- a/llvm/lib/Target/ARM/ARMInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMInstrInfo.cpp
@@ -25,7 +25,7 @@
 using namespace llvm;
 
 ARMInstrInfo::ARMInstrInfo(const ARMSubtarget &STI)
-  : RI(*this, STI), Subtarget(STI) {
+  : ARMBaseInstrInfo(STI), RI(*this, STI) {
 }
 
 unsigned ARMInstrInfo::getUnindexedOpcode(unsigned Opc) const {